1
0
Fork 0
mirror of https://github.com/Ryujinx/Ryujinx.git synced 2024-10-01 12:30:00 +02:00
Ryujinx/Ryujinx.Graphics.Shader/StructuredIr/StructuredProgramContext.cs

303 lines
8.7 KiB
C#
Raw Normal View History

New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
using Ryujinx.Graphics.Shader.IntermediateRepresentation;
2019-10-13 08:02:07 +02:00
using Ryujinx.Graphics.Shader.Translation;
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
using System.Collections.Generic;
using System.Linq;
using static Ryujinx.Graphics.Shader.StructuredIr.AstHelper;
namespace Ryujinx.Graphics.Shader.StructuredIr
{
class StructuredProgramContext
{
private HashSet<BasicBlock> _loopTails;
private Stack<(AstBlock Block, int EndIndex)> _blockStack;
private Dictionary<Operand, AstOperand> _localsMap;
private Dictionary<int, AstAssignment> _gotoTempAsgs;
private List<GotoStatement> _gotos;
private AstBlock _currBlock;
private int _currEndIndex;
public StructuredProgramInfo Info { get; }
2019-10-13 08:02:07 +02:00
public ShaderConfig Config { get; }
public StructuredProgramContext(int blocksCount, ShaderConfig config)
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
{
_loopTails = new HashSet<BasicBlock>();
_blockStack = new Stack<(AstBlock, int)>();
_localsMap = new Dictionary<Operand, AstOperand>();
_gotoTempAsgs = new Dictionary<int, AstAssignment>();
_gotos = new List<GotoStatement>();
_currBlock = new AstBlock(AstBlockType.Main);
_currEndIndex = blocksCount;
Info = new StructuredProgramInfo(_currBlock);
2019-10-13 08:02:07 +02:00
Config = config;
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
}
public void EnterBlock(BasicBlock block)
{
while (_currEndIndex == block.Index)
{
(_currBlock, _currEndIndex) = _blockStack.Pop();
}
if (_gotoTempAsgs.TryGetValue(block.Index, out AstAssignment gotoTempAsg))
{
AddGotoTempReset(block, gotoTempAsg);
}
LookForDoWhileStatements(block);
}
public void LeaveBlock(BasicBlock block, Operation branchOp)
{
LookForIfStatements(block, branchOp);
}
private void LookForDoWhileStatements(BasicBlock block)
{
// Check if we have any predecessor whose index is greater than the
// current block, this indicates a loop.
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
bool done = false;
foreach (BasicBlock predecessor in block.Predecessors.OrderByDescending(x => x.Index))
{
if (predecessor.Index < block.Index)
{
break;
}
if (predecessor.Index < _currEndIndex && !done)
{
Operation branchOp = (Operation)predecessor.GetLastOp();
NewBlock(AstBlockType.DoWhile, branchOp, predecessor.Index + 1);
_loopTails.Add(predecessor);
done = true;
}
else
{
AddGotoTempReset(block, GetGotoTempAsg(block.Index));
break;
}
}
}
private void LookForIfStatements(BasicBlock block, Operation branchOp)
{
if (block.Branch == null)
{
return;
}
bool isLoop = block.Branch.Index <= block.Index;
if (block.Branch.Index <= _currEndIndex && !isLoop)
{
NewBlock(AstBlockType.If, branchOp, block.Branch.Index);
}
else if (!_loopTails.Contains(block))
{
AstAssignment gotoTempAsg = GetGotoTempAsg(block.Branch.Index);
IAstNode cond = GetBranchCond(AstBlockType.DoWhile, branchOp);
AddNode(Assign(gotoTempAsg.Destination, cond));
AstOperation branch = new AstOperation(branchOp.Inst);
AddNode(branch);
GotoStatement gotoStmt = new GotoStatement(branch, gotoTempAsg, isLoop);
_gotos.Add(gotoStmt);
}
}
private AstAssignment GetGotoTempAsg(int index)
{
if (_gotoTempAsgs.TryGetValue(index, out AstAssignment gotoTempAsg))
{
return gotoTempAsg;
}
AstOperand gotoTemp = NewTemp(VariableType.Bool);
gotoTempAsg = Assign(gotoTemp, Const(IrConsts.False));
_gotoTempAsgs.Add(index, gotoTempAsg);
return gotoTempAsg;
}
private void AddGotoTempReset(BasicBlock block, AstAssignment gotoTempAsg)
{
AddNode(gotoTempAsg);
// For block 0, we don't need to add the extra "reset" at the beginning,
// because it is already the first node to be executed on the shader,
// so it is reset to false by the "local" assignment anyway.
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
if (block.Index != 0)
{
Info.MainBlock.AddFirst(Assign(gotoTempAsg.Destination, Const(IrConsts.False)));
}
}
private void NewBlock(AstBlockType type, Operation branchOp, int endIndex)
{
NewBlock(type, GetBranchCond(type, branchOp), endIndex);
}
private void NewBlock(AstBlockType type, IAstNode cond, int endIndex)
{
AstBlock childBlock = new AstBlock(type, cond);
AddNode(childBlock);
_blockStack.Push((_currBlock, _currEndIndex));
_currBlock = childBlock;
_currEndIndex = endIndex;
}
private IAstNode GetBranchCond(AstBlockType type, Operation branchOp)
{
IAstNode cond;
if (branchOp.Inst == Instruction.Branch)
{
cond = Const(type == AstBlockType.If ? IrConsts.False : IrConsts.True);
}
else
{
cond = GetOperandUse(branchOp.GetSource(0));
Instruction invInst = type == AstBlockType.If
? Instruction.BranchIfTrue
: Instruction.BranchIfFalse;
if (branchOp.Inst == invInst)
{
cond = new AstOperation(Instruction.LogicalNot, cond);
}
}
return cond;
}
public void AddNode(IAstNode node)
{
_currBlock.Add(node);
}
public GotoStatement[] GetGotos()
{
return _gotos.ToArray();
}
private AstOperand NewTemp(VariableType type)
{
AstOperand newTemp = Local(type);
Info.Locals.Add(newTemp);
return newTemp;
}
public AstOperand GetOperandDef(Operand operand)
{
if (TryGetUserAttributeIndex(operand, out int attrIndex))
{
Info.OAttributes.Add(attrIndex);
}
return GetOperand(operand);
}
public AstOperand GetOperandUse(Operand operand)
{
if (TryGetUserAttributeIndex(operand, out int attrIndex))
{
Info.IAttributes.Add(attrIndex);
2019-10-13 08:02:07 +02:00
Info.InterpolationQualifiers[attrIndex] = operand.Interpolation;
}
else if (operand.Type == OperandType.Attribute && operand.Value == AttributeConsts.InstanceId)
{
Info.UsesInstanceId = true;
New shader translator implementation (#654) * Start implementing a new shader translator * Fix shift instructions and a typo * Small refactoring on StructuredProgram, move RemovePhis method to a separate class * Initial geometry shader support * Implement TLD4 * Fix -- There's no negation on FMUL32I * Add constant folding and algebraic simplification optimizations, nits * Some leftovers from constant folding * Avoid cast for constant assignments * Add a branch elimination pass, and misc small fixes * Remove redundant branches, add expression propagation and other improvements on the code * Small leftovers -- add missing break and continue, remove unused properties, other improvements * Add null check to handle empty block cases on block visitor * Add HADD2 and HMUL2 half float shader instructions * Optimize pack/unpack sequences, some fixes related to half float instructions * Add TXQ, TLD, TLDS and TLD4S shader texture instructions, and some support for bindless textures, some refactoring on codegen * Fix copy paste mistake that caused RZ to be ignored on the AST instruction * Add workaround for conditional exit, and fix half float instruction with constant buffer * Add missing 0.0 source for TLDS.LZ variants * Simplify the switch for TLDS.LZ * Texture instructions related fixes * Implement the HFMA instruction, and some misc. fixes * Enable constant folding on UnpackHalf2x16 instructions * Refactor HFMA to use OpCode* for opcode decoding rather than on the helper methods * Remove the old shader translator * Remove ShaderDeclInfo and other unused things * Add dual vertex shader support * Add ShaderConfig, used to pass shader type and maximum cbuffer size * Move and rename some instruction enums * Move texture instructions into a separate file * Move operand GetExpression and locals management to OperandManager * Optimize opcode decoding using a simple list and binary search * Add missing condition for do-while on goto elimination * Misc. fixes on texture instructions * Simplify TLDS switch * Address PR feedback, and a nit
2019-04-18 01:57:08 +02:00
}
else if (operand.Type == OperandType.ConstantBuffer)
{
Info.CBuffers.Add(operand.GetCbufSlot());
}
return GetOperand(operand);
}
private AstOperand GetOperand(Operand operand)
{
if (operand == null)
{
return null;
}
if (operand.Type != OperandType.LocalVariable)
{
return new AstOperand(operand);
}
if (!_localsMap.TryGetValue(operand, out AstOperand astOperand))
{
astOperand = new AstOperand(operand);
_localsMap.Add(operand, astOperand);
Info.Locals.Add(astOperand);
}
return astOperand;
}
private static bool TryGetUserAttributeIndex(Operand operand, out int attrIndex)
{
if (operand.Type == OperandType.Attribute)
{
if (operand.Value >= AttributeConsts.UserAttributeBase &&
operand.Value < AttributeConsts.UserAttributeEnd)
{
attrIndex = (operand.Value - AttributeConsts.UserAttributeBase) >> 4;
return true;
}
else if (operand.Value >= AttributeConsts.FragmentOutputColorBase &&
operand.Value < AttributeConsts.FragmentOutputColorEnd)
{
attrIndex = (operand.Value - AttributeConsts.FragmentOutputColorBase) >> 4;
return true;
}
}
attrIndex = 0;
return false;
}
}
}